High Quality High Dynamic Range Imaging

A high dynamic range (HDR) image uses a large bit depth up to 32-bit per pixel per color channel. Due to hardware limitation, it can neither be captured by conventional camera in a single photo, nor be displayed on a conventional monitor. In this thesis, a software solution of HDR imaging is introduced from the synthesis to the display of high quality HDR images. Contrary to conventional imaging method, HDR imaging uses multiple exposures. Therefore, the variance of the same pixel in different images is non-linear and difficult to measure. To solve this, a new inter-pixel relationship function (IRF) is proposed using both spatial and temporal correlations. It is widely used in many HDR imaging components. Multiple input images produce a new ghosting artifact due to moving object. A real time de-ghosting method is first proposed using bi-directional comparison and IRF based moving object detection and patching. It is lightweight in terms of both time complexity and physical memory consumption, which makes it suitable for mobile devices. We further extend it by merging the IRF with a histogram intensity mapping and adopting a new threshold model based on statistical study. The extended deghosting scheme is more robust in the scenarios where the moving object occupies larger areas. Noise in the input images will be kept in the synthesized HDR image. A 2D de-noising factor is proposed for the synthesis of 32-bit HDR image, and a noise reduced tonemapping is proposed to map the 32-bit HDR image back to 8-bit for the display based on information content weighting (ICW). Both methods have low time complexity. It is also important to measure if the input images are from the same scene, as huge artifacts may appear if a wrong image is involved in the synthesis process. We proposed a structure similarly based metric which is robust to the images with different exposures. We also proposed a metric for the measurement of two 32-bit HDR images based on their histograms of radiance maps. AMS Classification: 68U10; 94A08 Universal Decimal Classification: 004.932 Library of Congress Subject Headings: High dynamic range imaging; Computational photography; Image enhancement Methods; Image systems Image quality Yleinen suomalainen asiasanasto: digitaalikuvaus; kuvankäsittely